[Playtest Feedback] First-Time Non-Technical User — Full Session Report (19 Issues + Strategic Recommendations)

[efalk524-ops]

Playtester Context

Tester: Emily Falk
Sessions: March 31 – April 1, 2026 (2 days)
Background: First-time user, no software development background
Project used: MTG Deck Builder (conversational AI deck builder with live card data, playtesting simulator, and multi-source pricing)

---

What Was Completed

• ✅ Full install and bootstrap (with AI assistance)
• ✅ Phase 1 — USER.md (user preferences)
• ✅ Phase 2 — PROJECT.md (MTG Deck Builder project)
• ✅ Research phase — libraries, APIs, pitfalls
• ✅ Full specification interview (9 questions)
• ✅ PRD reviewed, refined, and approved (22 FRs + 11 NFRs)
• ✅ SPECIFICATION.md reviewed and approved (8 phases, 38 tasks)
• 🔴 Build phase — NOT STARTED (cost/budget blocker)

---

Executive Summary

The product works. The interview-driven planning process is genuinely impressive and produces professional-quality output. However, the experience revealed a consistent and significant gap: Deft Directive is currently designed for experienced developers, and non-technical users encounter friction at nearly every decision point that could be resolved with plain English explanations, better onboarding, and one missing feature: cost/budget transparency before the build phase.

"I REALLY WANT to build this app — but I have financial realities I have to consider, and I have no idea what this will cost me."

---

Key Finding: AI Assistance Dependency

The most important finding of this playtest: a non-technical user can successfully complete the Deft Directive planning flow, but cannot do it without an AI assistant alongside them.

Throughout both sessions, I had an AI assistant available to explain terminology, recommend answers to technical questions, catch missing features before PRD approval, correct mistakes, and provide confidence to proceed at moments of uncertainty.

Without that assistance, I would have been blocked at the first technical question (frontend framework selection) and would not have produced a complete or accurate specification.

Recommendation: Consider a built-in "co-pilot mode" for non-technical users — an embedded AI assistant that explains, advises, and validates alongside the interview process. This would make Deft Directive truly self-contained for any user, regardless of background.

---

Issues Log

Install

#	Issue
1	Warp terminal not detected/installed — no prompt informing user it is required
2	Install path troubleshooting needed — if the desired install path (e.g. C:\) is not accessible, the installer fails silently with no guidance. Installing to the user folder (C:\Users\YourName\) works correctly. Clearer error messaging and path guidance would help non-technical users resolve this without assistance.
3	Git PATH requires full system restart — no in-app prompt to restart

Bootstrap

#	Issue
4	Agent picks up wrong AGENTS.md (Warp's own) — requires full path specification
5	Describing a project directly skips interview — agent jumps to build mode instead
15	New conversation pre-fills previous prompt — confusing for new users
18	Onboarding should suggest using a personal/neutral project — not work-adjacent
19	Work-adjacent project during playtesting raised ethical/conflict of interest concern — onboarding guidance needed

Interview

#	Issue
6	No error recovery or undo for wrong menu selections — blindly accepted
8	Technical framework questions (frontend, database, auth, deployment) presented to non-technical users without plain English explanation
16	Numbered menu options intermittently don't accept single digit input
17	Unexplained technical confirmations after selections (e.g. "Next.js — pairs with Vercel AI SDK")

PRD Review

#	Issue
9	"PRD" jargon unexplained — non-technical users don't know what it means
10	Accept/Refine/Edit button labels not intuitive — should use plain English or numbered menu
11	No clear approval action after PRD review — no button, no suggested input, no numbered menu
12	Red/green diff view alarming for non-technical users — "that looks scary, I thought I did something wrong"

Spec Review

#	Issue
12	Red/green diff view alarming (same issue, also present at spec review stage)
13	AI assistance prompt missing at SPECIFICATION.md review — most critical moment for non-technical users

Build

#	Issue
7	[CRITICAL BLOCKER] No cost/budget transparency before build phase — hard blocker for non-technical users, small businesses, and non-profits
14	"Would you like to start building?" missing pre-flight checklist (accounts needed, costs, timeline, step-by-step vs auto)

---

Critical Gap — Intellectual Property Not Flagged

At no point during the research, interview, or specification process did Deft Directive flag that Magic: The Gathering is intellectual property owned by Wizards of the Coast. Legal specifications were provided for gameplay rules and deck building strategies, and the full spec was approved — without a single mention of IP risk, copyright, trademark, or the need for legal disclaimers.

What Was Missed

The generated specification never:

• Asked whether the app was for personal or commercial use (monetization changes the legal picture entirely)
• Flagged that card images and oracle text are WotC copyright
• Suggested including a legal disclaimer in the app
• Recommended against hosting card assets directly on the developer's own servers
• Raised trademark concerns about using "Magic: The Gathering" in the app name or UI
• Suggested consulting a lawyer before building a commercial app around established IP

Risk Spectrum (independently researched)

• 🟢 Low risk: Gameplay engine + user-imported data only
• 🟡 Medium risk: API-based deck builder using Scryfall — where most community apps operate; tolerated but not legally safe
• 🔴 High risk: Hosting card assets directly + monetizing — significantly increases WotC's likelihood of action

Recommendation

When a project involves established intellectual property (games, music, film, books, sports leagues, etc.), Deft should automatically flag IP considerations during the research or interview phase:

• Plain English summary of IP risks for the specific domain
• Questions about monetization intent (free vs. commercial)
• Minimum protection recommendations (disclaimers, API-only asset access, hosting policy)
• Recommendation to consult a lawyer for any commercial application built on third-party IP

This applies far beyond MTG — it is a missing layer of due diligence for any project built around existing IP.

---

Strategic Recommendations

1. The Untapped Market: Non-Technical Users

Deft Directive currently targets experienced developers. But this playtest demonstrates that non-technical users — beginners, entrepreneurs, small businesses, non-profits — can complete the planning flow with support. This is an enormous untapped market. The value proposition ("anything you can dream of, built from a conversation") is extraordinary for this audience — but only if the product meets them where they are.

2. Tiered User Experience (Not a Tiered Product)

The product doesn't need to change. The language and guidance does. Suggested onboarding questions at first install:

• Is this for personal use or business use?
• What is your technical skill level? (Beginner / Intermediate / Expert)
• Do you have a budget in mind?

These three questions alone allow Deft to tailor the entire experience without changing the underlying technology. Simplifying the language doesn't reduce the sophistication of the product — a pilot and a passenger are on the same plane.

3. Pre-Build Cost & Budget Interview

Between specification approval and "start building," there should be a dedicated phase covering: estimated hosting costs, required accounts and API fees, monthly operating costs, and scale considerations. This is the natural moment — after the user knows what they're building but before they commit to building it. Without it, the product's value stops at specification for budget-conscious users.

4. Built-In AI Co-Pilot for Non-Technical Users

The single most impactful improvement: embed contextual AI assistance into the interview and review flow for non-technical users. Not a separate product — a mode within Deft that explains, advises, and validates alongside the process. This closes the gap that currently requires users to have a separate AI system running in another window.

---

Positive Observations

• The interview-driven planning process is genuinely impressive — one question at a time, well-paced
• The Research phase was a standout feature — searching the web and producing real, actionable technical guidance before the interview
• The PRD and SPECIFICATION output are professional quality — a real developer could build from them
• The product correctly sized the project, identified the biggest engineering risk (playtesting simulator), and flagged open questions proactively
• The vBRIEF tracking system is elegant — silent, automatic, no user burden
• The "just pick good defaults" option is exactly right — more of this philosophy throughout

---

Submitted by Emily Falk — First-Time Non-Technical Playtester | April 1–2, 2026

[MScottAdams]

Roadmap Refresh Analysis (2026-04-03)

Category: Playtest feedback / meta-issue (umbrella)

Summary: Comprehensive playtest report from a non-technical first-time user. Completed the full planning flow (install → bootstrap → interview → PRD → SPECIFICATION) but hit a hard stop before build due to cost/budget transparency. Logs 19 issues across install, bootstrap, interview, PRD review, spec review, and build phases. Also identifies a critical gap: no IP/legal risk flagging. Includes 4 strategic recommendations.

Relationship to existing issues:

• Warp doesn't load deft's AGENTS.md by default (but I have a work around) #136 (Phase 2) — Warp doesn't load AGENTS.md by default; playtest issue make /deft read-only #4 (wrong AGENTS.md) is same root cause
• feat: Allow users to change technical rating (1/2/3) when starting a new project #77 (Phase 5) — Allow users to change technical rating; playtest recommendation deft/run output clean up #2 (tiered UX by skill level) aligns directly
• Deft as teacher: contract hierarchy, explain WHY, and adaptive teaching mode #84 (Phase 2/5) — Deft as teacher / adaptive teaching; core finding is non-technical users need plain English explanations
• Deft identity and positioning: resolve naming before README reframe #89 (Phase 2) — Identity and positioning; playtest suggests untapped non-technical user market

Scope: Large — umbrella issue with 19 sub-issues and 4 strategic recommendations. Individual items range from trivial (wording) to major (cost/budget interview, IP risk flagging).

Recommendation: Phase 2 as tracked umbrella. Content/wording fixes (jargon, plain English, diff view guidance) fit Phase 2. Strategic recommendations (cost interview, co-pilot mode, tiered UX) are Phase 5 inputs. IP risk flagging is a novel finding worth a dedicated issue later.